[zeromq-dev] Application design. Overcomplicated?

Ilja Golshtein ilejncs at narod.ru
Fri Apr 8 10:04:17 CEST 2011


Hi Martin

08.04.2011, 09:44, "Martin Sustrik" <sustrik at 250bpm.com>:

>>  My question is what is wrong in the design above and what could be simplified.
>
> So what's the scenario? A bunch of stateless workers accessing the same
> physical storage?

Yes.

It would be nice to create a generic Unix 0mq based multithread C application design.
A sort of best practice.

This application
* terminates upon a signal
* understands several different commands and process these command in parallel returning some results.

My particular questions are
* is it Ok to have poll in every worker 
* is it Ok to have one socket for one command coming from outside

Main consideration - performance.

Some of important issues like "Making a Clean Exit" are covered in Pieter's Zguide (which is great), though
I think an example might be useful.

> In such a case simple request/reply pattern should do. Maybe an
> additional pub/sub feed to distribute cluster-scoped commands to all the
> workers.

You mean there is no reason to do multithreading at all? Often it is true,
but let's focus on cases when it makes sense.

Thanks.

-- 
Best regards,
Ilja Golshtein.



More information about the zeromq-dev mailing list